Over the past several months, we’ve explored the transformative potential of artificial intelligence from multiple angles. First, we laid the groundwork by describing what AI is and how it may be poised to reshape the way we work and live. We then turned our focus to where AI is already showing up in markets and portfolios, highlighting key inputs in the buildout and the best ways for investors to gain exposure to this emerging investment theme. Most recently, we took a step back to consider AI through a macroeconomic lens, comparing it to past waves of innovation and examining its potentially transformative impact on labor markets, GDP, and productivity.
Now, as AI adoption continues to rapidly accelerate, we shift our attention to the current market environment. In this installment of our AI blog series, we will provide an update on the magnitude of capital expenditures being funneled into AI infrastructure, evaluate the sustainability of the AI capex cycle, and conclude by addressing the most important question of all: Is this the rational exuberance of a technological revolution, or are we beginning to see signs of speculative excess?
An Update on the AI Spending Boom
One of the core principles of a capitalistic system is the idea that prices, profits, and losses serve as signals that guide investment decisions, helping to ensure that capital is deployed where it can generate the greatest value. Implicit in this system is the idea that businesses are relatively efficient allocators of capital, but this concept is now being put to the test in the context of the AI infrastructure boom.
The amount of capital being directed into AI-related investments is staggering, with consensus estimates indicating that total capex over the next five years will be measured in the trillions. The magnitude of such spending suggests that the major hyperscalers (companies such as Amazon, Microsoft, Alphabet, and Meta) view this as a pivotal moment, whereby establishing a dominant position is existential to their businesses.
With so much at stake, companies are moving quickly to gain early-mover advantages. At this stage of the AI buildout, the focus is on securing the computing power required to train advanced models and prepare them for commercial use. As we discussed in a previous blog, AI demands an extraordinary amount of computational resources, and these demands are so intense that computing power is rapidly becoming one of the world’s most valuable commodities. The estimated scale of the related infrastructure buildout is illustrated in the chart below:
The Earnings Feedback Loop
This intense focus on computing power and the associated surge in capital expenditure has prompted some to question the sustainability of the cycle, with particular concerns being expressed about the circular nature of many of these investments. Some companies are investing huge sums in their own customers, who then spend that money on the investor’s products – creating a circular flow of funds. One example of this sort of activity is NVIDIA’s recent pledge to invest $100 billion in OpenAI, with OpenAI committed to building data centers that run primarily on NVIDIA’s GPUs.
Such interlocked funding deals improve the financial forecasts for all parties involved, as AI startups announce infrastructure plans and suppliers report surging sales. This has led to speculation about whether such activity is artificially inflating earnings and masking the emergence of an AI-fueled asset bubble. However, not everyone thinks that circular funding is inherently bad. Some market analysts argue that this is business as usual in capital-intensive industries, especially in cases such as AI Infrastructure where the prohibitively high costs of such investments limit the number of participants who can afford a seat at the table.
There is also historical precedent for such activity, and it has been a relatively common practice in industries like telecom, aviation, and energy for many decades. In the case of NVIDIA, the massive stockpile of cash on its balance sheet gives the company latitude to become a strategic financier to the AI industry, almost like a bank for its customers. The strategy is to give NVIDIA a stake in the AI ecosystem beyond just selling chips.
On the other hand, the scale and optics of these arrangements have raised concern. Critics argue that these deals create “phantom revenue”. Essentially, a company may report growing sales, but those sales were facilitated by money that the company itself provided to the customer.
Figure: Many critics argue that the circuitous nature of AI investment spending is creating a fragile system that is vulnerable to external shocks. Source: ChatGPT
Another worry is that such tightly interwoven deals make the whole system fragile. The agreements tie the fate of multiple companies together, where if one party stumbles, the rest fall with it. Imagine OpenAI fails to monetize its new supercomputer capacity. It might then default on payments it promised to NVIDIA. Then Nvidia’s expected revenues don’t materialize, and the company fails to meet the market’s earnings expectations. The circuitous cycle then begins to spiral downwards.
Market Implications
So, should investors be concerned? In our view, the answer is both yes and no. On one hand, cynicism is a valuable trait for disciplined investors to possess, and the circuitous nature of the AI investment cycle poses unique concentration risks to the market. The risk that revenue and earnings forecasts are overly optimistic is exacerbated by the fact that the fates of many prominent companies are tied together. But on the other hand, many of the major players are already incredibly stable and profitable businesses with healthy balance sheets. In most cases, the AI spending we discussed above is being funded from the operating cash flow that these companies are generating, enabling them to make big commitments without jeopardizing their core businesses (or even their obligations to each other). If we eventually move into a new phase of speculation where companies are taking on debt and levering their balance sheets in an effort to “keep up,” that would signal to us that we have entered a new stage of risk taking that begins to mirror previous periods of economic excess. But to this point, that has not yet happened at scale.
The chart below helps to illustrate this point. While some companies are certainly spending more than others, none of the below hyperscalers are overextended in terms of their capital allocation decisions relative to operating cash flow.
The AI industry today stands at a crossroads between innovation and speculation. Corporate leaders increasingly view success in the AI arms race as a critical strategic priority, which has accelerated both the scale and pace of infrastructure investment. On the surface, this wave of spending appears more disciplined than past bubbles, as it is largely funded by cash flows rather than leverage. However, the circular nature of these investments introduces a layer of fragility through mutual dependence.
Ultimately, whether this moment becomes a launchpad for long-term value creation or a cautionary tale of speculative excess will depend on the ability of AI technologies to deliver real, scalable economic impact. Early indicators suggest that the major hyperscalers are already seeing strong ROI, with many reporting notable productivity gains without proportional head-count growth (a concept we covered in the third installment of this series). In fact, a recent research piece from Goldman Sachs found that some early adopters are experiencing productivity improvements of 25-30% following the deployment of AI applications.1 If these results can be replicated at scale, the resulting expansion in profit margins and economic value could more than justify the upfront costs. At that point, the circularity of the capex cycle would be viewed not as a vulnerability, but as an important foundation that helped facilitate a technological revolution.
Concluding Thoughts
Like with all nascent technologies, there are likely to be winners and losers, and it’s unclear whether today’s AI leaders will retain their dominant positions. For now, the market appears to be rewarding early adopters, but sustained success will require real economic value creation on a massive scale. Investors should recognize the tremendous potential going forward but remain cognizant of risks and vulnerabilities in the market. In our view, the best way to invest is to embrace long-term growth potential while recognizing that the path forward will almost certainly not be linear.
[1] Goldman Sachs Research. “The AI Spending Boom is not Too Big.” October 2025
Wealthspire Advisors LLC and certain of its affiliates are separately registered investment advisers. © 2025 Wealthspire
This material should not be construed as a recommendation, offer to sell, or solicitation of an offer to buy a particular security or investment strategy. The information provided is for informational purposes only and should not be relied upon for accounting, legal, or tax advice. While the information is deemed reliable, Wealthspire cannot guarantee its accuracy, completeness, or suitability for any purpose, and makes no warranties with regard to the results to be obtained from its use.
